一种改进的图像三维重建技术研究
|
摘要
目前图像三维重建得到了广泛应用。摄像机成像仪上的光条纹中心是实现三维重建的重要信息源。针对光条纹中心提取方法中的Steger算法计算量大、运算时间较长的缺陷,本文将GPU多网格(grid)多块(block)多线程(thread)的并行运算优势,与Steger算法每个像素点卷积运算相互独立的特点相结合,设计了一种利用GPU加速的改进的Steger算法,以减少运算耗时。实验结果表明:GPU下的光条纹提取时间为CPU下的1/16到1/12,大大提高了运算效率,减少了计算时间。
Presently three-dimensional reconstruction of images has been used and the extraction of optical fringe center of imaging instrument in camera is very important information source in the procedure of 3 D reconstruction.Aimed to the shortage of huge calculation and long operation time of Steger algorithm,which is a famous method,combining the advantage of GPU with multi-grid,multi-block and multi-thread with characteristic of Steger algorithm's the two-dimensional Gaussian convolution with GPU acceleration of pixel respectively,this paper preposed an improved Steger algorithm by using GPU acceleration to reduce the operation time. The experiment results showed extraction time of optical fringe center in GPU is 1/16 to 1/12 that of CPU.
Presently three-dimensional reconstruction of images has been used and the extraction of optical fringe center of imaging instrument in camera is very important information source in the procedure of 3 D reconstruction.Aimed to the shortage of huge calculation and long operation time of Steger algorithm,which is a famous method,combining the advantage of GPU with multi-grid,multi-block and multi-thread with characteristic of Steger algorithm's the two-dimensional Gaussian convolution with GPU acceleration of pixel respectively,this paper preposed an improved Steger algorithm by using GPU acceleration to reduce the operation time. The experiment results showed extraction time of optical fringe center in GPU is 1/16 to 1/12 that of CPU.
引文
[1]尚超.序列图像的三维重建技术研究[D].西安:西北大学,2012.
[2]朱梦楠.三维扫描中激光光条中心提取精度的研究[D].咸阳:西北农林科技大学,2017.
[3]苏小勤,熊显名.快速线结构光中心提取算法[J].计算机应用,2016,36(1):238-242.
[4]许宁.线结构光光条图像处理方法研究[D].哈尔滨:哈尔滨工程大学,2007.
[5]STEGER C.An unbiased detector of curvilinear structures[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2002,20(2):113-125.
[6]李栋梁,唐阳山,黄贤成,等.基于Hessian矩阵的线结构光中心线提取方法研究[J].汽车使用技术,2017,(22):37-39.
[7]袁明.基于多激光条纹的手持式快速三维测量关键技术[D].武汉:华中科技大学,2015.
[8]李凤娇,李小菁,刘震.基于多尺度分析的激光光条中心点坐标提取方法[J].光学学报,2014,34(11):103-108.
[9]LUO Gongzhi,YANG Xiaojiang,ZHOU Dequn.Rough analysis model of multi-attribute decision making based on limited extended dominance relation[J].Journal of Systems&Management,2009,18(4):391-396.
[10]贾卫平,王邦国.基于Hessian矩阵的多结构光条纹中心快速提取方法[J].大连大学学报,2014,(35)6:34-37.
[11]岳龙,郭喜庆,余锦.自适应复杂光条纹中心提取方法研究[J].科学技术与工程,2016,16(14):236-239.
[12]秦子扬,夏桂锁,伏燕军,等.基于差影法的Hessian矩阵激光条纹中心提取方法研究[J].南昌航空大学学报:自然科学版,2017,(31)03:28-33.
[13]PRESS W,TEUKOLSKY S,VETTERLING W,FLAN-NERY B.Numerical recipes in C:the art of scientific computing,1988[M].Cambridge,MA:Press Syndiacate University of Cambridge.
[14]HARTLEY R I.Projective reconstruction and invariants from multiple images[J].Pattern Analysis and Machine Intelligence,IEEE Transactions on,1994,16(10):1036-1041.
[2]朱梦楠.三维扫描中激光光条中心提取精度的研究[D].咸阳:西北农林科技大学,2017.
[3]苏小勤,熊显名.快速线结构光中心提取算法[J].计算机应用,2016,36(1):238-242.
[4]许宁.线结构光光条图像处理方法研究[D].哈尔滨:哈尔滨工程大学,2007.
[5]STEGER C.An unbiased detector of curvilinear structures[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2002,20(2):113-125.
[6]李栋梁,唐阳山,黄贤成,等.基于Hessian矩阵的线结构光中心线提取方法研究[J].汽车使用技术,2017,(22):37-39.
[7]袁明.基于多激光条纹的手持式快速三维测量关键技术[D].武汉:华中科技大学,2015.
[8]李凤娇,李小菁,刘震.基于多尺度分析的激光光条中心点坐标提取方法[J].光学学报,2014,34(11):103-108.
[9]LUO Gongzhi,YANG Xiaojiang,ZHOU Dequn.Rough analysis model of multi-attribute decision making based on limited extended dominance relation[J].Journal of Systems&Management,2009,18(4):391-396.
[10]贾卫平,王邦国.基于Hessian矩阵的多结构光条纹中心快速提取方法[J].大连大学学报,2014,(35)6:34-37.
[11]岳龙,郭喜庆,余锦.自适应复杂光条纹中心提取方法研究[J].科学技术与工程,2016,16(14):236-239.
[12]秦子扬,夏桂锁,伏燕军,等.基于差影法的Hessian矩阵激光条纹中心提取方法研究[J].南昌航空大学学报:自然科学版,2017,(31)03:28-33.
[13]PRESS W,TEUKOLSKY S,VETTERLING W,FLAN-NERY B.Numerical recipes in C:the art of scientific computing,1988[M].Cambridge,MA:Press Syndiacate University of Cambridge.
[14]HARTLEY R I.Projective reconstruction and invariants from multiple images[J].Pattern Analysis and Machine Intelligence,IEEE Transactions on,1994,16(10):1036-1041.